Real-Time Recognition Algorithm of Small Target for UAV Infrared Detection

Qianqian Zhang,Li Zhou,Junshe An

DOI: https://doi.org/10.3390/s24103075

IF: 3.9

2024-05-13

Sensors

Abstract:Unmanned Aerial Vehicle (UAV) infrared detection has problems such as weak and small targets, complex backgrounds, and poor real-time detection performance. It is difficult for general target detection algorithms to achieve the requirements of a high detection rate, low missed detection rate, and high real-time performance. In order to solve these problems, this paper proposes an improved small target detection method based on Picodet. First, to address the problem of poor real-time performance, an improved lightweight LCNet network was introduced as the backbone network for feature extraction. Secondly, in order to solve the problems of high false detection rate and missed detection rate due to weak targets, the Squeeze-and-Excitation module was added and the feature pyramid structure was improved. Experimental results obtained on the HIT-UAV public dataset show that the improved detection model's real-time frame rate increased by 31 fps and the average accuracy (MAP) increased by 7%, which proves the effectiveness of this method for UAV infrared small target detection.

engineering, electrical & electronic,instruments & instrumentation,chemistry, analytical

What problem does this paper attempt to address?

This paper aims to address the issue of small target detection in UAV (Unmanned Aerial Vehicle) infrared detection, specifically including the following aspects:

1. **Weak and Small Targets and Complex Backgrounds**: The targets in UAV infrared images are usually small and weak in intensity, with complex backgrounds, making detection more difficult.
2. **Poor Real-time Performance**: Existing general object detection algorithms struggle to achieve high real-time performance while ensuring high detection rates and low miss rates.
3. **High False Detection and Miss Rates**: Due to the inconspicuous features of the targets and the coexistence of multiple scales, existing methods tend to have high false detection and miss rates.

To address these issues, the authors propose an improved small target detection method based on the Picodet network:
- Using an improved lightweight LCNet network as the backbone network to extract features, enhancing real-time performance.
- Introducing the Squeeze-and-Excitation module and improving the feature pyramid structure to enhance the detection capability for weak and small targets.
- Experimental results show that on the HIT-UAV public dataset, the improved model achieves a real-time frame rate increase of 31 fps and an average precision (MAP) increase of 7%, demonstrating the effectiveness of the proposed method.